Embedded human control of robots using myoelectric interfaces

Chris Wilson Antuvan, Mark Ison, Panagiotis Artemiadis

Research output: Contribution to journalArticle

31 Citations (Scopus)

Abstract

Myoelectric controlled interfaces have become a research interest for use in advanced prostheses, exoskeletons, and robot teleoperation. Current research focuses on improving a user's initial performance, either by training a decoding function for a specific user or implementing "intuitive" mapping functions as decoders. However, both approaches are limiting, with the former being subject specific, and the latter task specific. This paper proposes a paradigm shift on myoelectric interfaces by embedding the human as controller of the system to be operated. Using abstract mapping functions between myoelectric activity and control actions for a task, this study shows that human subjects are able to control an artificial system with increasing efficiency by just learning how to control it. The method efficacy is tested by using two different control tasks and four different abstract mappings relating upper limb muscle activity to control actions for those tasks. The results show that all subjects were able to learn the mappings and improve their performance over time. More interestingly, a chronological evaluation across trials reveals that the learning curves transfer across subsequent trials having the same mapping, independent of the tasks to be executed. This implies that new muscle synergies are developed and refined relative to the mapping used by the control task, suggesting that maximal performance may be achieved by learning a constant, arbitrary mapping function rather than dynamic subject- or task-specific functions. Moreover, the results indicate that the method may extend to the neural control of any device or robot, without limitations for anthropomorphism or human-related counterparts.

Original languageEnglish (US)
Article number6720133
Pages (from-to)820-827
Number of pages8
JournalIEEE Transactions on Neural Systems and Rehabilitation Engineering
Volume22
Issue number4
DOIs
StatePublished - 2014

Fingerprint

Robots
Learning
Muscles
Learning Curve
Research
Upper Extremity
Prostheses and Implants
Muscle
Efficiency
Equipment and Supplies
Remote control
Prosthetics
Decoding
Controllers
Transfer (Psychology)

Keywords

  • Electromyography
  • human-robot interaction
  • motor learning
  • myoelectric control
  • real-time systems

ASJC Scopus subject areas

  • Neuroscience(all)
  • Computer Science Applications
  • Biomedical Engineering
  • Medicine(all)

Cite this

Embedded human control of robots using myoelectric interfaces. / Antuvan, Chris Wilson; Ison, Mark; Artemiadis, Panagiotis.

In: IEEE Transactions on Neural Systems and Rehabilitation Engineering, Vol. 22, No. 4, 6720133, 2014, p. 820-827.

Research output: Contribution to journalArticle

@article{15647952b806419cbb0ee4bd5f31ab59,
title = "Embedded human control of robots using myoelectric interfaces",
abstract = "Myoelectric controlled interfaces have become a research interest for use in advanced prostheses, exoskeletons, and robot teleoperation. Current research focuses on improving a user's initial performance, either by training a decoding function for a specific user or implementing {"}intuitive{"} mapping functions as decoders. However, both approaches are limiting, with the former being subject specific, and the latter task specific. This paper proposes a paradigm shift on myoelectric interfaces by embedding the human as controller of the system to be operated. Using abstract mapping functions between myoelectric activity and control actions for a task, this study shows that human subjects are able to control an artificial system with increasing efficiency by just learning how to control it. The method efficacy is tested by using two different control tasks and four different abstract mappings relating upper limb muscle activity to control actions for those tasks. The results show that all subjects were able to learn the mappings and improve their performance over time. More interestingly, a chronological evaluation across trials reveals that the learning curves transfer across subsequent trials having the same mapping, independent of the tasks to be executed. This implies that new muscle synergies are developed and refined relative to the mapping used by the control task, suggesting that maximal performance may be achieved by learning a constant, arbitrary mapping function rather than dynamic subject- or task-specific functions. Moreover, the results indicate that the method may extend to the neural control of any device or robot, without limitations for anthropomorphism or human-related counterparts.",
keywords = "Electromyography, human-robot interaction, motor learning, myoelectric control, real-time systems",
author = "Antuvan, {Chris Wilson} and Mark Ison and Panagiotis Artemiadis",
year = "2014",
doi = "10.1109/TNSRE.2014.2302212",
language = "English (US)",
volume = "22",
pages = "820--827",
journal = "IEEE Transactions on Neural Systems and Rehabilitation Engineering",
issn = "1534-4320",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "4",

}

TY - JOUR

T1 - Embedded human control of robots using myoelectric interfaces

AU - Antuvan, Chris Wilson

AU - Ison, Mark

AU - Artemiadis, Panagiotis

PY - 2014

Y1 - 2014

N2 - Myoelectric controlled interfaces have become a research interest for use in advanced prostheses, exoskeletons, and robot teleoperation. Current research focuses on improving a user's initial performance, either by training a decoding function for a specific user or implementing "intuitive" mapping functions as decoders. However, both approaches are limiting, with the former being subject specific, and the latter task specific. This paper proposes a paradigm shift on myoelectric interfaces by embedding the human as controller of the system to be operated. Using abstract mapping functions between myoelectric activity and control actions for a task, this study shows that human subjects are able to control an artificial system with increasing efficiency by just learning how to control it. The method efficacy is tested by using two different control tasks and four different abstract mappings relating upper limb muscle activity to control actions for those tasks. The results show that all subjects were able to learn the mappings and improve their performance over time. More interestingly, a chronological evaluation across trials reveals that the learning curves transfer across subsequent trials having the same mapping, independent of the tasks to be executed. This implies that new muscle synergies are developed and refined relative to the mapping used by the control task, suggesting that maximal performance may be achieved by learning a constant, arbitrary mapping function rather than dynamic subject- or task-specific functions. Moreover, the results indicate that the method may extend to the neural control of any device or robot, without limitations for anthropomorphism or human-related counterparts.

AB - Myoelectric controlled interfaces have become a research interest for use in advanced prostheses, exoskeletons, and robot teleoperation. Current research focuses on improving a user's initial performance, either by training a decoding function for a specific user or implementing "intuitive" mapping functions as decoders. However, both approaches are limiting, with the former being subject specific, and the latter task specific. This paper proposes a paradigm shift on myoelectric interfaces by embedding the human as controller of the system to be operated. Using abstract mapping functions between myoelectric activity and control actions for a task, this study shows that human subjects are able to control an artificial system with increasing efficiency by just learning how to control it. The method efficacy is tested by using two different control tasks and four different abstract mappings relating upper limb muscle activity to control actions for those tasks. The results show that all subjects were able to learn the mappings and improve their performance over time. More interestingly, a chronological evaluation across trials reveals that the learning curves transfer across subsequent trials having the same mapping, independent of the tasks to be executed. This implies that new muscle synergies are developed and refined relative to the mapping used by the control task, suggesting that maximal performance may be achieved by learning a constant, arbitrary mapping function rather than dynamic subject- or task-specific functions. Moreover, the results indicate that the method may extend to the neural control of any device or robot, without limitations for anthropomorphism or human-related counterparts.

KW - Electromyography

KW - human-robot interaction

KW - motor learning

KW - myoelectric control

KW - real-time systems

UR - http://www.scopus.com/inward/record.url?scp=84904330475&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84904330475&partnerID=8YFLogxK

U2 - 10.1109/TNSRE.2014.2302212

DO - 10.1109/TNSRE.2014.2302212

M3 - Article

VL - 22

SP - 820

EP - 827

JO - IEEE Transactions on Neural Systems and Rehabilitation Engineering

JF - IEEE Transactions on Neural Systems and Rehabilitation Engineering

SN - 1534-4320

IS - 4

M1 - 6720133

ER -